As we are using YOLOv8, there are likely some documentation on how to deploy a nano model to something like a Jetson Nano. Probably conversion to TensorRT.
Check the note in utils/jetson/README.md on the Jetson Nano setup.
YOLOv8n model is in this full dataset Dropbox link.
Salmon counting code is in training/pysalmcount/pysalmcount/salmon_counter.py